ea83e1c05a
At the moment, the full cost of all interleave group members is assigned to the instruction at the group's insert position, even if the decision was to not form an interleave group. This can lead to inaccurate cost estimates, e.g. if the instruction at the insert position is dead. If the decision is to not vectorize but scalarize or scather/gather, then the cost will be to total cost for all members. In those cases, assign individual the cost per member, to more closely reflect to choice per instruction. This fixes a divergence between legacy and VPlan-based cost model. Fixes https://github.com/llvm/llvm-project/issues/108098.
44 lines
2.2 KiB
LLVM
44 lines
2.2 KiB
LLVM
; REQUIRES: asserts
|
|
; RUN: opt -mtriple=s390x-unknown-linux -mcpu=z16 -passes=loop-vectorize \
|
|
; RUN: -debug-only=loop-vectorize -force-vector-width=4 \
|
|
; RUN: -disable-output < %s 2>&1 | FileCheck %s
|
|
;
|
|
; Check cost function for <8 x i128> store interleave group.
|
|
|
|
; CHECK: LV: Checking a loop in 'fun'
|
|
; CHECK: LV: Found an estimated cost of 4 for VF 4 For instruction: store i128 8721036757475490113
|
|
; CHECK: LV: Found an estimated cost of 4 for VF 4 For instruction: store i128 8721036757475490113
|
|
|
|
define noundef i32 @fun(i32 %argc, ptr nocapture readnone %argv) {
|
|
entry:
|
|
%l_4774.i = alloca [4 x [2 x i128]], align 8
|
|
call void @llvm.lifetime.start.p0(i64 128, ptr nonnull %l_4774.i)
|
|
br label %for.cond4.preheader.i
|
|
|
|
for.cond4.preheader.i: ; preds = %for.cond4.preheader.i, %entry
|
|
%indvars.iv8.i = phi i64 [ 0, %entry ], [ %indvars.iv.next9.i, %for.cond4.preheader.i ]
|
|
%arrayidx10.i = getelementptr inbounds [4 x [2 x i128]], ptr %l_4774.i, i64 0, i64 %indvars.iv8.i, i64 0
|
|
store i128 8721036757475490113, ptr %arrayidx10.i, align 8
|
|
%arrayidx10.i.c = getelementptr inbounds [4 x [2 x i128]], ptr %l_4774.i, i64 0, i64 %indvars.iv8.i, i64 1
|
|
store i128 8721036757475490113, ptr %arrayidx10.i.c, align 8
|
|
%indvars.iv.next9.i = add nuw nsw i64 %indvars.iv8.i, 1
|
|
%exitcond.not.i = icmp eq i64 %indvars.iv.next9.i, 4
|
|
br i1 %exitcond.not.i, label %func_1.exit, label %for.cond4.preheader.i
|
|
|
|
func_1.exit: ; preds = %for.cond4.preheader.i
|
|
%arrayidx195.i = getelementptr inbounds [4 x [2 x i128]], ptr %l_4774.i, i64 0, i64 1
|
|
%0 = load i128, ptr %arrayidx195.i, align 8
|
|
%cmp200.i = icmp ne i128 %0, 0
|
|
%conv202.i = zext i1 %cmp200.i to i64
|
|
%call203.i = tail call i64 @safe_sub_func_int64_t_s_s(i64 noundef %conv202.i, i64 noundef 9139899272418802852)
|
|
call void @llvm.lifetime.end.p0(i64 128, ptr nonnull %l_4774.i)
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.cond, %func_1.exit
|
|
br label %for.cond
|
|
}
|
|
|
|
declare void @llvm.lifetime.start.p0(i64, ptr nocapture)
|
|
declare void @llvm.lifetime.end.p0(i64, ptr nocapture)
|
|
declare dso_local i64 @safe_sub_func_int64_t_s_s(i64, i64)
|